Altered Multi-Voxel Prefrontal And Mesolimbic Patterns Associated With Reward Processing In Schizophrenia: Evidence From Representational Similarity Analysis

Background: Accumulating evidence has shown that patients with schizophrenia exhibit dysfunctional prefrontal and mesolimbic responses to anticipation and receipt of reward. However, conventional univariate analysis employed in previous studies involved considerable smoothing and averaging approach that might lead to sensitivity reduction for detecting subtle changes in functionally heterogeous areas. Representational Similarity Analysis (RSA), which is one type of multi-voxel pattern analysis (MVPA), may overcome this limitation by capturing fine-grained changes involved in the reward processing. In the present study, we adopted the RSA to examine whether patients with schizophrenia exhibited altered multivariate activation pattern associated reward processing. Methods: We recruited 33 patients with schizophrenia and 29 age and gender matched healthy controls. All the participants were asked to undertake the Food Incentive Delay (FID) task inside a scanner. Food images (not real food) were used as reinforcement stimuli in receipt phase. Activation maps for each individual valence condition (including positive, neutral, negative and baseline) during each reward processing phase (anticipation, response, receipt) were used as input for searchlight-based multi-voxel pattern analyses. RSA was used to identify regions associated with models of phase and valence as well as group differences in correlation maps. Multiple comparisons were adjusted with Family-wise correction as P < .05 at cluster level. Results: Patients with schizophrenia showed lower correlations between the model reflecting differentiation of value processing among 3 reward phases and multi-voxel response patterns at the left middle frontal gyrus (MFG, x/yx/z: −61/19/48; t = 5.29, cluster size =69) and the right putamen at a marginal significance level (x/yx/z: −33/−3/3; t = 4.38, cluster size = 50, P FWE-cluster level = .075) compared to healthy controls. Furthermore, schizophrenia patients also exhibited weaker correlation between the model indicating value differentiation among positive, neutral, negative and baseline conditions and multi-voxel activation patterns mainly at the left orbitofrontal cortex (x/yx/z: −27/54/−12; t = 4.63, cluster size =48), the left dorsolateral prefrontal cortex (x/yx/z: −21/54/36; t = 4.37, cluster size = 62) and the brainstem (x/yx/z: −6/−39/−30; t = 4.49, cluster size =43). Conclusion: These findings suggest that prefrontal and mesolimbic regions play an important role in dysfunctional reward processing in schizophrenia.